Mapping microphytobenthos in the intertidal zone of Northern ...

Chl a concentrations in the supernatant have then been converted in terms of Chl a per surface unit (Chl a, mg.m -2 )
taking into account the surface of the sampling units.
2.3. Reflectance spectra analysis
In the field reflectance spectra, Chl a absorption is located around 673 nm which is a different wavelength position
than what is published for pure pigment (665 nm, [37], [38]). This is probably due to the fact that this wavelength
position is for pure pigment in a solvent and not in a leaving cell. Some tests we performed in the laboratory with
monospecific cultures also show an absorption located around 673 nm. The importance of the absorption is directly
related to pigment concentration and was estimated using three different and simple approaches (Figure 2).
2.3.1. Single band ratio (Figure 2a)
This approach is a method that has been classically used in spectroscopy and remote sensing in general to measure
amounts of atmospheric or surface components. It is based on differential absorption concept ([39], [40]) which
consists of a simple ratio between reflectance at maximum absorption (Rb) and reflectance outside the absorption,
referred to as the continuum (Rc). Here we used a ratio between reflectance at 673 nm (absorption) and reflectance
at 720 nm (continuum).
2.3.2. Normalized ratio (or scaled band depth) (Figure 2b, 2c)
Since the objective is to develop a simple method that has to be independent from measurement conditions, one
needs to perform some kind of normalization in order to remove the influence of other parameters on the spectral
signature and concentrate on the chlorophyll absorption itself. For example, it is well known that sediment grain
size influences the diffusing part of light, therefore changing the general shape of the spectrum through a
modification of the spectral dependency of the diffusion ([41], [42]). Spectral shape is also influenced by sediment
optical properties such as refracting index related to sediment composition. Sediment moisture content will also
modify the spectral signature (enhanced liquid water absorptions, level change of the general reflectance).
Removing the continuum of the spectrum allows isolating the spectral feature from other effects. This approach was
proposed by Clark and Roush [43] who determined the depth of a spectral absorption feature and related it to frost
grain size and by Clark [44] and Clark et al. [45] and references therein for applications to rocks and minerals. The
scaled band depth Db is calculated as the difference between the continuum reflectance Rc and the reflectance
spectrum Rb in the deepest part of the absorption band, normalized by the continuum reflectance:
Db = (Rc – Rb) / Rc. (2)
While this band depth method is a valuable approach to the problem, its reliance on a single band causes the
accuracy of the results to suffer in the presence of noise in the reflectance spectrum. Noise-induced changes in the
spectrum affect the depth of the absorption feature and may give erroneous chlorophyll concentration estimates.
2.3.3. Integral after continuum removal (scaled band area) (Figure 2d)
To minimize any effects of instrumental noise on chlorophyll retrievals, we also tested the use of the scaled area of
the absorption feature Ab rather than simply the scaled absorption band depth. This approach was used by Nolin and
Dozier [46] to estimate snow grain size from hyperspectral data. Scaled band area is a dimensionless quantity and is
calculated by integrating the scaled absorption band depth over the wavelengths of the absorption feature:
Ab = ∫λ (Rc – Rb) / Rc. (3)
The basic assumption is that the noise is randomly distributed Gaussian noise (“white noise”). While the exact
distribution of sensor noise is not known, a normal distribution is a reasonable assumption and has been used by
others in examining sensor noise characteristics ([47]). By integrating over the absorption feature, the fluctuations
caused by noise should average out and produce an estimate closer to the true value. Integration was restricted to
the [650 – 720 nm] wavelength range in order to exclude influence of other absorbing pigments such as Chlorophyll
c (absorption at 590 and 635 nm).
398